This is the spine of the stack: the model whose output is what we sell against the bookmaker. Player ratings ultimately feed into here; market mechanics define the target metric. The job of this note is to map the modelling families, who built them, what the strongest public baselines are, and where bottom-up player approaches start to add value.
Modeling
Team Success and Match Outcome Modeling — Landscape Note (Area B)
Summary + sample · full document is 1,931 words
Summary
Sample
The two modelling traditions
Goal-process models treat scoring as a stochastic process — typically Poisson — where each team has a latent attack/defence rate. Results are simulated by sampling goals.
Outcome-prediction models skip the goal process and directly output P(home / draw / away) — or P(cover) / P(over) for AH and totals — from features. Discriminative classifiers, ratings systems, and regression models all sit here.
The two traditions complement each other. Goal models give you a coherent joint distribution over scores (good for derived markets: BTTS, correct score, AH lines), at the cost of strong distributional assumptions. Discriminative models often beat goal models on raw 1X2 log-loss but can't price exotic markets coherently.
Full document
Pro
Want the full document?
Team Success and Match Outcome Modeling — Landscape Note (Area B) runs 1,931 words. Pro members get every research note in full, plus the arbitrage feed, model outputs, and weekly updates.